Data-driven subtypes of mild cognitive impairment and their modifiable determinants: latent profile analysis of the BIOCIS cohort

Abstract

Mild cognitive impairment (MCI) is heterogenous, but classification often overlooks regional variability. Cross-sectional data including 764 individuals from the Biomarker and Cognition study, Singapore, aged >50 years with MCI based on Montreal Cognitive Assessment normative cut-offs were utilised. Latent profile analyses utilising BIC thresholds were applied to demographically adjusted cognitive assessments. Hierarchical regression evaluated demographic, cardiometabolic, psychological, lifestyle and biomarker predictors of subtypes. Participants had a mean age of 62.72 ± 7.66, 59% females, 14.50 ± 3.44 education years. Four subtypes of MCI emerged; above-average performers (39.0%), multi-domain impaired (19.6%), visuospatial preserved (15.2%), and processing speed preserved (26.2%). Relative to multi-domain impaired, lower body mass index, alcohol consumption and better quality of life were associated with above-average performers and processing speed preserved membership. Greater grip strength and lower cardiometabolic burden were associated with visuospatial preserved membership. Plasma Aβ40 showed a preliminary, exploratory association with cluster membership. In conclusion, the study identified four cognitively derived subtypes of MCI, with strong associations observed for lifestyle and psychological factors. Hence, highlighting the potential value of personalised, early approaches to MCI management.